using 
CSV,
Dates,
DataFrames,
Format,
Markdown,
Measures,
Plots,
RCall,
Statistics,
StatsPlots,
Suppressor
@suppress R"library(forecast)"
@suppress gr()
;

death_path = "COVID-19/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_global.csv"
confirmed_path = "COVID-19/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv"
;

death = CSV.read(death_path) |> DataFrame! |> x -> rename!(x, Dict(Symbol("Province/State") => :state, Symbol("Country/Region") => :country))
confirmed = CSV.read(confirmed_path) |> DataFrame! |> x -> rename!(x, Dict(Symbol("Province/State") => :state, Symbol("Country/Region") => :country))
;

"""
    convert an array of columns and column names into a markdown table

    parameters: 
        cols : Array[Array[Any]] | DataFrame
            an array of arrays which are the columns of the table
        names : Array[String]
            an array of column names. If inputting a df, input `string.(names(df))`
        df : Boolean
            whether or not the cols are an array of arrays or a dataframe

    returns: String
        string of markdown text
"""
function markdown_table(cols, names; df=false)
    if df
        cols = [col for col in eachcol(cols)]
    end
    output_string = ""
    cols = [repr.(col) for col in cols]
    for i in 1:length(names)
        output_string = output_string * "|" * names[i]
    end
    output_string *= "\n"
    for i in 1:length(names)
        output_string = output_string * "|" * "---"
    end
    output_string *= "\n"
    for row in 1:length(cols[1])
        for col in 1:length(cols)
            output_string = output_string * "|" * cols[col][row]
        end
        output_string *= "\n"
    end
    # displaying strings without quotes
    output_string =  replace(output_string, "\"" => "")

    return output_string
end
;

function country_data(country; state=false, start_date=Dates.Date(2020, 1, 22), end_date="max")
    aggregate(x) = [sum(col) for col in eachcol(x)]
    
    state_mask = typeof(state) == String ? death.state .== state : true
    d = death[(death.country .== country) .& state_mask, 5:end] |> aggregate
    dates = [Dates.Date(2020, 1, 22) + Dates.Day(day) for day in 1:length(d)] |> x -> reshape(x, :, 1)
    state_mask = typeof(state) == String ? confirmed.state .== state : true
    c = confirmed[(confirmed.country .== country) .& state_mask, 5:end] |> aggregate
    
    df = convert(DataFrame, dates) |> x -> rename!(x, [:date])
    country_name = typeof(state) == String ? country * "-" * state : country
    df[!, :country] .= country_name
    df[!, :confirmed] = c
    df[!, :deaths] = d
    df[!, :death_rate] = d ./ c
    df[!, :new_cases] .= 0
    df[2:end, :new_cases] = df[2:end, :confirmed] .- df[1:end-1, :confirmed]
    df[!, :new_deaths] .= 0
    df[2:end, :new_deaths] = df[2:end, :deaths] .- df[1:end-1, :deaths]
    df[!, :acceleration_cases] .= 0
    df[2:end, :acceleration_cases] = df[2:end, :new_cases] .- df[1:end-1, :new_cases]
    df[!, :acceleration_deaths] .= 0
    df[2:end, :acceleration_deaths] = df[2:end, :new_deaths] .- df[1:end-1, :new_deaths]
    df[!, :days_since_100] .= 0
    counter = 0
    for row in eachrow(df)
        if row.confirmed > 100
            counter += 1
        end
        row.days_since_100 = counter
    end
    df[!, :days_since_10] .= 0
    counter = 0
    for row in eachrow(df)
        if row.deaths > 10
            counter += 1
        end
        row.days_since_10 = counter
    end
    
    if end_date == "max"
        end_date = maximum(df.date)
    end
    return df[(df.date .>= start_date) .& (df.date .<= end_date), :]
end
;

function plot_country(country; state=false, metric=:confirmed, start_date=Dates.Date(2020, 1, 22), end_date="max", days_since_100=false)
    data = country_data(country; state=state, start_date=start_date, end_date=end_date)
    if days_since_100
        if metric == :confirmed
            plot(data[data.days_since_100 .> 0, :days_since_100], data[data.days_since_100 .> 0, metric], label=country, legend=:outertopright, size=(1000, 500))
        else
            plot(data[data.days_since_10 .> 0, :days_since_10], data[data.days_since_10 .> 0, metric], label=country, legend=:outertopright, size=(1000, 500))
        end
    else
        plot(data.date, data[!, metric], label=country, legend=:outertopright, size=(1000, 500))
    end
    
    plot!([0], linetype=:hline, color=:black, label="")
    plot!(yformatter=:plain)
end
function plot_country!(country; state=false, metric=:confirmed, start_date=Dates.Date(2020, 1, 22), end_date="max", days_since_100=false)
    data = country_data(country; state=state, start_date=start_date, end_date=end_date)
    if days_since_100
        if metric == :confirmed
            plot!(data[data.days_since_100 .> 0, :days_since_100], data[data.days_since_100 .> 0, metric], label=country, legend=:outertopright, size=(1000, 500))
        else
            plot!(data[data.days_since_10 .> 0, :days_since_10], data[data.days_since_10 .> 0, metric], label=country, legend=:outertopright, size=(1000, 500))
        end
    else
        plot!(data.date, data[!, metric], label=country, legend=:outertopright, size=(1000, 500))
    end
    plot!(yformatter=:plain)
end
bar_plot(country, metric) = bar(
    all_country_data[
        (all_country_data.country .== country) .& (all_country_data.date .>= Dates.Date(2020, 1, 1)),
        :date
    ], 
    all_country_data[
        (all_country_data.country .== country) .& (all_country_data.date .>= Dates.Date(2020, 1, 1)),
        metric
    ], 
    legend=false, 
    linecolor=1, 
    title=country,
    yformatter=:plain
)
function accel_plot(country_name="World")
    if country_name == "World"
        tmp = all_country_data
    else
        tmp = all_country_data[all_country_data.country .== country_name, :]
    end
    agg = by(tmp, [:date], :confirmed => sum, :deaths => sum)
    conf = plot(agg.date, agg.confirmed_sum, legend=false, title="$country_name Confirmed Cases", size=(1400, 700), color=:grey, yformatter=:plain)
    death_plot = plot(agg.date, agg.deaths_sum, legend=false, title="$country_name Deaths", size=(1400, 700), color=:orange, yformatter=:plain)

    # derivates
    agg[!, :new_cases] .= 0
    agg[!, :new_cases][2:end] = agg.confirmed_sum[2:end] - agg.confirmed_sum[1:end-1]
    agg[!, :new_deaths] .= 0
    agg[!, :new_deaths][2:end] = agg.deaths_sum[2:end] - agg.deaths_sum[1:end-1]
    agg[!, :conf_acceleration] .= 0
    agg[!, :conf_acceleration][2:end] = agg.new_cases[2:end] - agg.new_cases[1:end-1]
    agg[!, :death_acceleration] .= 0
    agg[!, :death_acceleration][2:end] = agg.new_deaths[2:end] - agg.new_deaths[1:end-1]

    # moving average
    moving_average(vs,n) = [sum(@view vs[i:(i+n-1)])/n for i in 1:(length(vs)-(n-1))]
    n_days = 3
    ma_date = agg.date[1 + n_days-1:end]
    conf_slope = moving_average(agg.new_cases, n_days)
    death_slope = moving_average(agg.new_deaths, n_days)
    conf_acc = moving_average(agg.conf_acceleration, n_days)
    death_acc = moving_average(agg.death_acceleration, n_days)

    p_conf_slope = plot(ma_date, conf_slope, legend=false, title="$country_name 1st Derivative (new cases)", size=(1400, 700), color=:grey)
    plot!([0], linetype=:hline, color=:black, label="", yformatter=:plain)

    p_death_slope = plot(ma_date, death_slope, legend=false, title="$country_name 1st Derivative (new deaths)", size=(1400, 700), color=:orange)
    plot!([0], linetype=:hline, color=:black, label="", yformatter=:plain)

    p_conf_acc = plot(ma_date, conf_acc, legend=false, title="$country_name 2nd Derivative (cases acceleration)", size=(1400, 700), color=:grey)
    plot!([0], linetype=:hline, color=:black, label="", yformatter=:plain)

    p_death_acc = plot(ma_date, death_acc, legend=false, title="$country_name 2nd Derivative (deaths acceleration)", size=(1400, 700), color=:orange)
    plot!([0], linetype=:hline, color=:black, label="", yformatter=:plain)

    plot(conf, death_plot, p_conf_slope, p_death_slope, p_conf_acc, p_death_acc, layout=(3,2))
end
;

function r_forecast(x, y; r_forecast_function, time_function, h)    
    fc_x = [x[end] + time_function(i) for i in 1:h]
    fc = r_forecast_function(y)
    point_forecast = [val for val in fc[2]]
    lo_80 = [val for val in fc[6][1:h]]
    hi_80 = [val for val in fc[5][1:h]]
    lo_95 = [val for val in fc[6][h+1:h*2]]
    hi_95 = [val for val in fc[5][h+1:h*2]]
    output = hcat([point_forecast, lo_80, hi_80, lo_95, hi_95]...) |> DataFrame |> x -> rename!(x, [:point_forecast, :lo_80, :hi_80, :lo_95, :hi_95])

    p = plot(x, y, legend=false, label="Observed", color=:black, formatter=:plain)
    plot!(fc_x, point_forecast, ribbon=(point_forecast.-lo_95, hi_95.-point_forecast), label="", color=:lightblue, fillalpha=0.5)
    plot!(fc_x, point_forecast, ribbon=(point_forecast.-lo_80, hi_80.-point_forecast), label="Forecast", color=:blue, fillalpha=0.3)
    
    return [fc, output, p]
end
;

function individual_country(country; comparison_country="Italy")
    # line plots
    cases_100 = plot_country(country, days_since_100=true, metric=:confirmed)
    plot_country!(comparison_country, days_since_100=true, metric=:confirmed)
    plot!(title="$country Days Since 100th Case", titlefont=font(10))
    deaths_10 = plot_country(country, days_since_100=true, metric=:deaths)
    plot_country!(comparison_country, days_since_100=true, metric=:deaths)
    plot!(title="$country Days Since 10th Death", titlefont=font(10))
    line_plots = plot(cases_100, deaths_10, layout=(2,1))
    
    #accel plots
    accel_plots = accel_plot(country)
    plot!(titlefont=font(10))
    
    # bar plots
    cases_bar = bar_plot(country, :new_cases)
    plot!(title="$country New Cases Per Day", titlefont=font(10))
    deaths_bar = bar_plot(country, :new_deaths)
    plot!(title="$country Deaths Per Day", titlefont=font(10))
    bar_plots = plot(cases_bar, deaths_bar, layout=(2,1))
    
    # forecast plots
    data = country_data(country) |> x -> x[x.confirmed .>= 100,:]
    forecast_plot_cases = r_forecast(data.date, data.confirmed, r_forecast_function = z -> R"holt"(z, h=20, damped=true), time_function=Dates.Day, h=20)[3]
    plot!(title="$country Forecast Cases", titlefont=font(10))
    forecast_plot_deaths = r_forecast(data.date, data.deaths, r_forecast_function = z -> R"holt"(z, h=20, damped=true), time_function=Dates.Day, h=20)[3]
    plot!(title="$country Forecast Deaths", titlefont=font(10))
    forecast_plots = plot(forecast_plot_cases, forecast_plot_deaths, layout=(1, 2))
    
    final_plot = plot(accel_plots, line_plots, bar_plots, forecast_plots, layout=(4, 1))
    plot!(size=(900, 2000), margin=5mm)
    return final_plot
end
;

all_countries = death.country |> unique |> sort
all_country_data = vcat([country_data(country) for country in all_countries]...)
current_state = by(all_country_data, [:country], :confirmed => maximum, :deaths => maximum)
current_state[!, :death_rate] = current_state.deaths_maximum ./ current_state.confirmed_maximum
;

Markdown.parse("""
    # Overview of COVID-19 as of $(maximum(all_country_data.date))
    This page uses data from https://github.com/CSSEGISandData/COVID-19 (Johns Hopkins CSSE is the original source) to create a report on the status of COVID-19 cases and deaths around the world. It is updated once a day around 10:00am CET.
    """)

Worldwide Current State¶

tmp = current_state
Markdown.parse("""
- Confirmed cases: **$(format(sum(tmp.confirmed_maximum), commas=true))**
- Deaths: **$(format(sum(tmp.deaths_maximum), commas=true))**
- Death Rate: **$(round(sum(tmp.deaths_maximum) / sum(tmp.confirmed_maximum) * 100, digits=2))%**
""")

accel_plot("World")
plot!(size=(900, 700))

Top 10 Worst Affected Countries¶

Confirmed¶

top_10 = sort(current_state, order(:confirmed_maximum, rev=true))[1:10, :]
top_10[!, :death_rate] = string.(round.(top_10.death_rate * 100, digits=2)) .* "%"
Markdown.parse(
    markdown_table(top_10, ["Country", "Confirmed Cases", "Deaths", "Death Rate"]; df=true)
)

Acceleration (>1000 cases now)¶

The Acceleration of Last 5 Days column is calculated by the average second derivative over the last 5 days / number of cases 5 days ago. It doesn't have much intrinsic meaning but is rather a more comparable/relative measure between countries of how fast new cases are accelerating.

countries = unique(all_country_data.country)
acceleration = []
cases_5_ago = []
cases_now = []
for country in countries
    metric = mean(all_country_data[all_country_data.country .== country, :acceleration_cases][end-4:end])
    metric /= all_country_data[all_country_data.country .== country, :confirmed][end-4]
    if isnan(metric)
        metric = 0
    end
    push!(acceleration, metric)
    push!(cases_5_ago, all_country_data[all_country_data.country .== country, :confirmed][end-4])
    push!(cases_now, all_country_data[all_country_data.country .== country, :confirmed][end])
end
last_5 = rename!(DataFrame([countries, acceleration, cases_now, cases_5_ago]), [:country, :last_5_accel, :cases_now, :cases_5_ago])
last_5[!, :perc_increase] = last_5.cases_now ./ last_5.cases_5_ago .- 1
tmp = sort(last_5[(last_5.cases_5_ago .> 20) .& (last_5.cases_now .> 1000), :], order(:last_5_accel, rev=true))
tmp[!, :perc_increase] = string.(round.(tmp.perc_increase .* 100, digits=2)) .* "%"
rename!(tmp, [:Country, Symbol("Acceleration of Last 5 Days"), Symbol("Cases Now"), Symbol("Cases 5 Days Ago"), Symbol("% Increase in 5 Days")])
tmp[!, Symbol("Acceleration of Last 5 Days")] = string.(round.(tmp[!, Symbol("Acceleration of Last 5 Days")] * 100, digits=2)) .* "%"

Markdown.parse(
    markdown_table(tmp, string.(names(tmp)), df=true)
)

Countries¶

Days since 100th case and 10th death plotted with Italy for comparison. Only countries with >= 500 confirmed cases show graphs. Derivative plots are 3 days moving average.

Forecasts¶

Forecast numbers and plots are created using Holt's linear trend method with dampening. Important to note is that this is just one of many methods that can forecast the data, and I have not spent a significant amount of time validating it. I may spend more time in the future investigating better forecasting methods.
Forecast numbers are given for 10 and 20 days out with the 80% prediction interval. Plots display the point forecast and 80% prediction interval (dark blue shading), and 95% prediction interval (light blue shading).

for coun in all_countries
    data = country_data(coun)
    display(Markdown.parse("## $coun"))
    Markdown.parse("""
- Confirmed cases: **$(format(maximum(data.confirmed), commas=true))**
- Deaths: **$(format(maximum(data.deaths), commas=true))**
- Death Rate: **$(round(maximum(data.deaths) / maximum(data.confirmed) * 100, digits=2))%**
""") |> display
    if maximum(data.confirmed) >= 500 || coun == "Hungary" 
        h = 20
        raw_fc_data = data[data.confirmed .>= 100, :]
        fc_data_cases = r_forecast(raw_fc_data.date, raw_fc_data.confirmed, r_forecast_function = z -> R"holt"(z, h=h, damped=true), time_function=Dates.Day, h=h)[2]
        fc_data_cases = round.(fc_data_cases, digits=0)
        fc_data_deaths = r_forecast(raw_fc_data.date, raw_fc_data.deaths, r_forecast_function = z -> R"holt"(z, h=h, damped=true), time_function=Dates.Day, h=h)[2]
        fc_data_deaths = round.(fc_data_deaths, digits=0)
        _10_days = maximum(data.date) + Dates.Day(10)
        _20_days = maximum(data.date) + Dates.Day(20)
        Markdown.parse("""
**$coun Forecast**
<br>
_By $(_10_days)_
- Confirmed cases: **$(format(fc_data_cases[10, :point_forecast], commas=true))** (+/- **$(format(fc_data_cases[10, :point_forecast] - fc_data_cases[10, :lo_80], commas=true))**)
- Deaths: **$(format(fc_data_deaths[10, :point_forecast], commas=true))** (+/- **$(format(fc_data_deaths[10, :point_forecast] - fc_data_deaths[10, :lo_80], commas=true))**)
_By $(_20_days)_
- Confirmed cases: **$(format(fc_data_cases[20, :point_forecast], commas=true))** (+/- **$(format(fc_data_cases[20, :point_forecast] - fc_data_cases[20, :lo_80], commas=true))**)
- Deaths: **$(format(fc_data_deaths[20, :point_forecast], commas=true))** (+/- **$(format(fc_data_deaths[20, :point_forecast] - fc_data_deaths[20, :lo_80], commas=true))**)
""") |> display
        display(individual_country(coun))
    end
end

┌ Warning: RCall.jl: Warning in ets(x, "AAN", alpha = alpha, beta = beta, phi = phi, damped = damped,  :
│   Not enough data to use damping
└ @ RCall /home/danhopp/.julia/packages/RCall/g7dhB/src/io.jl:113
┌ Warning: RCall.jl: Warning in ets(x, "AAN", alpha = alpha, beta = beta, phi = phi, damped = damped,  :
│   Not enough data to use damping
└ @ RCall /home/danhopp/.julia/packages/RCall/g7dhB/src/io.jl:113
┌ Warning: RCall.jl: Warning in ets(x, "AAN", alpha = alpha, beta = beta, phi = phi, damped = damped,  :
│   Not enough data to use damping
└ @ RCall /home/danhopp/.julia/packages/RCall/g7dhB/src/io.jl:113
┌ Warning: RCall.jl: Warning in ets(x, "AAN", alpha = alpha, beta = beta, phi = phi, damped = damped,  :
│   Not enough data to use damping
└ @ RCall /home/danhopp/.julia/packages/RCall/g7dhB/src/io.jl:113

Country	Confirmed Cases	Deaths	Death Rate
US	308850	8407	2.72%
Spain	126168	11947	9.47%
Italy	124632	15362	12.33%
Germany	96092	1444	1.5%
France	90848	7574	8.34%
China	82543	3330	4.03%
Iran	55743	3452	6.19%
United Kingdom	42477	4320	10.17%
Turkey	23934	501	2.09%
Switzerland	20505	666	3.25%

Country	Acceleration of Last 5 Days	Cases Now	Cases 5 Days Ago	% Increase in 5 Days
France	8.02%	90848	52827	71.97%
United Arab Emirates	6.02%	1505	664	126.66%
Japan	5.35%	3139	1953	60.73%
Qatar	4.89%	1325	781	69.65%
India	4.12%	3082	1397	120.62%
Brazil	3.43%	10360	5717	81.21%
Ukraine	2.48%	1225	645	89.92%
Russia	2.4%	4731	2337	102.44%
Serbia	2.31%	1624	900	80.44%
Finland	2.19%	1882	1418	32.72%
Argentina	2.11%	1451	1054	37.67%
Turkey	2.07%	23934	13531	76.88%
Panama	1.86%	1673	1181	41.66%
US	1.31%	308850	188172	64.13%
Romania	1.21%	3613	2245	60.94%
Czechia	1.19%	4472	3308	35.19%
Estonia	1.13%	1039	745	39.46%
Egypt	1.07%	1070	710	50.7%
Peru	1.0%	1746	1065	63.94%
Colombia	0.95%	1406	906	55.19%
Belgium	0.94%	18431	12775	44.27%
United Kingdom	0.88%	42477	25481	66.7%
Denmark	0.87%	4269	3039	40.47%
Singapore	0.86%	1189	926	28.4%
Chile	0.83%	4161	2738	51.97%
Luxembourg	0.73%	2729	2178	25.3%
Mexico	0.6%	1688	1094	54.3%
Ecuador	0.53%	3465	2240	54.69%
Portugal	0.52%	10524	7443	41.39%
South Africa	0.5%	1585	1353	17.15%
Poland	0.44%	3627	2311	56.95%
Ireland	0.22%	4604	3235	42.32%
Algeria	0.2%	1251	716	74.72%
Italy	0.14%	124632	105792	17.81%
Pakistan	0.12%	2818	1938	45.41%
Norway	0.08%	5550	4641	19.59%
Greece	0.06%	1673	1314	27.32%
Germany	0.04%	96092	71808	33.82%
Korea, South	0.03%	10156	9786	3.78%
Netherlands	0.03%	16727	12667	32.05%
China	-0.01%	82543	82279	0.32%
Malaysia	-0.04%	3483	2766	25.92%
Sweden	-0.07%	6443	4435	45.28%
Israel	-0.09%	7851	5358	46.53%
Saudi Arabia	-0.18%	2179	1563	39.41%
Spain	-0.18%	126168	95923	31.53%
Iceland	-0.23%	1417	1135	24.85%
Switzerland	-0.23%	20505	16605	23.49%
Iran	-0.28%	55743	44605	24.97%
Indonesia	-0.3%	2092	1528	36.91%
Philippines	-0.5%	3094	2084	48.46%
Thailand	-0.57%	2067	1651	25.2%
Australia	-0.69%	5550	4559	21.74%
Croatia	-0.69%	1126	867	29.87%
Dominican Republic	-0.76%	1488	1109	34.17%
Austria	-1.13%	11781	10180	15.73%
Canada	-1.35%	12978	8527	52.2%

Overview of COVID-19 as of 2020-04-05¶

Worldwide Current State¶

Top 10 Worst Affected Countries¶

Confirmed¶

Acceleration (>1000 cases now)¶

Countries¶

Forecasts¶

Afghanistan¶

Albania¶

Algeria¶

Andorra¶

Angola¶

Antigua and Barbuda¶

Argentina¶

Armenia¶

Australia¶

Austria¶

Azerbaijan¶

Bahamas¶

Bahrain¶

Bangladesh¶

Barbados¶

Belarus¶

Belgium¶

Belize¶

Benin¶

Bhutan¶

Bolivia¶

Bosnia and Herzegovina¶

Botswana¶

Brazil¶

Brunei¶

Bulgaria¶

Burkina Faso¶

Burma¶

Burundi¶

Cabo Verde¶

Cambodia¶

Cameroon¶

Canada¶

Central African Republic¶

Chad¶

Chile¶

China¶

Colombia¶

Congo (Brazzaville)¶

Congo (Kinshasa)¶

Costa Rica¶

Cote d'Ivoire¶

Croatia¶

Cuba¶

Cyprus¶

Czechia¶

Denmark¶

Diamond Princess¶

Djibouti¶

Dominica¶

Dominican Republic¶

Ecuador¶

Egypt¶

El Salvador¶

Equatorial Guinea¶

Eritrea¶

Estonia¶

Eswatini¶

Ethiopia¶

Fiji¶

Finland¶

France¶

Gabon¶

Gambia¶

Georgia¶

Germany¶

Ghana¶

Greece¶

Grenada¶

Guatemala¶

Guinea¶

Guinea-Bissau¶

Guyana¶